The Network Coding Based Schemes And Algorithms For Data Recovery And Retransmission In Wireless Networks

Network Coding(NC) is one of the most promising techniques to improve throughput and reliability for a wireless network. This dissertation investigates issues of network opportunistic coding in a variety of wireless transmission environments, proposes a new scheme for packet buffering and feedbacking, and develops an opportunistic listening and coding algorithm based on the theory of Markov decision process. Thus the dissertation provides some effective solutions for maximizing coding efficiency used in data recovery for reliable wireless networking communication. The main researches and achievements are summarized as follows.(1) For unicast transmission, this dissertation proposes a new packet buffering and feedbacking scheme. This scheme requires receivers to store all overheard undecodable packets and periodically report the reception status to the sender. Based on this idea, this dissertation designs a new coding algorithm to XOR coded packets with native packets. This algorithm makes the sender to get another and more coding opportunities of the undecodable packets that would be ignored or discarded in the conventional NC approach. Simulation results show that the new method can reduce the retransmission numbers from 8% to 21% comparing to the conventional method.(2) For multicast transmission, this dissertation presents a new NC optimization model. This model considers the data retransmission process as a sequential decision-making process. Based on this model, this dissertation designs a MDP-based NC approach that equates the optimal coding decision problem with the dynamic programming problem. In order to reduce the complexity and improve real-time performance, this dissertation develops a forward search based online planning algorithm, which can reduce the number of iterations as well as the calculations in each iteration. Simulation results demonstrate that the proposed approach can reduce the retransmission numbers from 5% to 14% by taking into account the impact of different coding decision on transmission performance.(3) For limited feedback, this dissertation proposes a new coding opportunity prediction scheme. This scheme models the data retransmission process based on Partially Observable Markov Decision Process(POMDP) theory, and then use belief state to estimate the existing coding opportunities in networks. Based on this model, this dissertation designs a NC-based data recovery scheme with single-step forward algorithm. The scheme determines the optimal combinations of packets by making quantification of the impact of each coding decision on the coding opportunities. Simulation results demonstrate that compared with existing approach, the proposed scheme can reduce the retransmission numbers from 1% to 7% in wireless multicast networks with limited feedback.(4) For hybrid unicast and multicast communication with single source, this dissertation presents a new data recovery scheme. The proposed scheme allows the sender to simultaneously retransmit unicast’s and multicast’s data under certain conditions to overcome the limitation of exiting approaches, which independently handle unicast and multicast lost packets. Simulation results show that the proposed scheme can reduce the retransmission numbers from 20% to 45% by exploiting the coding opportunities between unicast and multicast flows.This dissertation thoroughly studies the application issues of network coding aiming to reliable wireless communication, sets up corresponding analytical optimization models and proposes several effective NC-based date recovery schemes for different wireless scenarios. The achievements of this research possess valuable theoretical foundations and practical algorithm solutions for issues such as to reduce the number of retransmissions, to lower the transmission delay and to increase transmission reliability.